The n e w e ng l a n d j o u r na l of m e dic i n e

Original Article

Early Rhythm-Control Therapy in Patients

with Atrial Fibrillation
P. Kirchhof, A.J. Camm, A. Goette, A. Brandes, L. Eckardt, A. Elvan, T. Fetsch,
I.C. van Gelder, D. Haase, L.M. Haegeli, F. Hamann, H. Heidbüchel,
G. Hindricks, J. Kautzner, K.-H. Kuck, L. Mont, G.A. Ng, J. Rekosz, N. Schoen,
U. Schotten, A. Suling, J. Taggeselle, S. Themistoclakis, E. Vettorazzi, P. Vardas,
K. Wegscheider, S. Willems, H.J.G.M. Crijns, and G. Breithardt,
for the EAST-AFNET 4 Trial Investigators*​​

A BS T R AC T
BACKGROUND
Despite improvements in the management of atrial fibrillation, patients with this The authors’ full names, academic de-
condition remain at increased risk for cardiovascular complications. It is unclear grees, and affiliations are listed in the Ap-
pendix. Address reprint requests to Dr.
whether early rhythm-control therapy can reduce this risk. Kirchhof at the Department of Cardiolo-
METHODS gy, University Heart and Vascular Center,
Universitätsklinikum Hamburg–Eppendorf
In this international, investigator-initiated, parallel-group, open, blinded-outcome- Hamburg, Martinistraße 52, Gebäude
assessment trial, we randomly assigned patients who had early atrial fibrillation Ost 70, 20246 Hamburg, Germany, or at
(diagnosed ≤1 year before enrollment) and cardiovascular conditions to receive either ­p​.­kirchhof@​­uke​.­de.
early rhythm control or usual care. Early rhythm control included treatment with *A complete list of the EAST-AFNET 4
antiarrhythmic drugs or atrial fibrillation ablation after randomization. Usual care investigators is provided in the Supple-
mentary Appendix, available at NEJM.org.
limited rhythm control to the management of atrial fibrillation–related symptoms.
The first primary outcome was a composite of death from cardiovascular causes, This article was published on August 29,
2020, at NEJM.org.
stroke, or hospitalization with worsening of heart failure or acute coronary syn-
drome; the second primary outcome was the number of nights spent in the hospital DOI: 10.1056/NEJMoa2019422
Copyright © 2020 Massachusetts Medical Society.
per year. The primary safety outcome was a composite of death, stroke, or serious
adverse events related to rhythm-control therapy. Secondary outcomes, including
symptoms and left ventricular function, were also evaluated.
RESULTS
In 135 centers, 2789 patients with early atrial fibrillation (median time since di-
agnosis, 36 days) underwent randomization. The trial was stopped for efficacy at
the third interim analysis after a median of 5.1 years of follow-up per patient. A
first-primary-outcome event occurred in 249 of the patients assigned to early rhythm
control (3.9 per 100 person-years) and in 316 patients assigned to usual care (5.0 per
100 person-years) (hazard ratio, 0.79; 96% confidence interval, 0.66 to 0.94; P = 0.005).
The mean (±SD) number of nights spent in the hospital did not differ significantly
between the groups (5.8±21.9 and 5.1±15.5 days per year, respectively; P = 0.23). The
percentage of patients with a primary safety outcome event did not differ signifi-
cantly between the groups; serious adverse events related to rhythm-control ther-
apy occurred in 4.9% of the patients assigned to early rhythm control and 1.4% of
the patients assigned to usual care. Symptoms and left ventricular function at 2 years
did not differ significantly between the groups.
CONCLUSIONS
Early rhythm-control therapy was associated with a lower risk of cardiovascular
outcomes than usual care among patients with early atrial fibrillation and cardio-
vascular conditions. (Funded by the German Ministry of Education and Research
and others; EAST-AFNET 4 ISRCTN number, ISRCTN04708680; ClinicalTrials.gov
number, NCT01288352; EudraCT number, 2010​-­021258​-­20.)

n engl j med  nejm.org 1

The New England Journal of Medicine
Downloaded from nejm.org by gaspar caponi on September 2, 2020. For personal use only. No other uses without permission.
Copyright © 2020 Massachusetts Medical Society. All rights reserved.
 The n e w e ng l a n d j o u r na l of m e dic i n e

E
ven with current guideline-based
management, patients with atrial fibrilla-
tion have stroke, acute coronary syndrome,
heart failure, and cardiovascular death at a rate
of approximately 5% of patients per year,1-4 and
35 to 50% of patients with atrial fibrillation who
receive adequate anticoagulation either receive
inpatient therapy or die within 5 years.5,6 These
complications occur even though most atrial fi-
brillation–related ischemic strokes can be pre-
vented with anticoagulation,3,4 and rate control
often renders patients asymptomatic.7,8 The risk
of cardiovascular complications is increased dur-
ing the first year after atrial fibrillation is diag-
nosed (a period referred to here as “early atrial
fibrillation”).9 Furthermore, rhythm-control ther-
apy may be more effective when delivered early.10,11
Previous trials, including one trial involving
patients with heart failure, have not shown su-
periority of rhythm control with antiarrhythmic
drugs over rate control in patients with estab-
lished atrial fibrillation.7,8,12,13 Small trials have
suggested that atrial fibrillation ablation may im-
prove left ventricular function and may reduce the
risk of adverse outcomes in patients with atrial
fibrillation and heart failure,2,14 and in one trial
the antiarrhythmic drug dronedarone, as com-
pared with placebo, reduced the composite out-
come of death and cardiovascular hospitaliza-
tions.6 Some reports have indicated low rates of
stroke and death associated with rhythm-control
therapy,5,15 including atrial fibrillation ablation.16-18
The Early Treatment of Atrial Fibrillation for
Stroke Prevention Trial (EAST-AFNET 4) therefore
was designed to test whether a strategy of early
rhythm-control therapy that includes atrial fi-
brillation ablation would be associated with bet-
ter outcomes in patients with early atrial fibril-
lation than contemporary, evidence-based usual
care.11
Me thods
Trial Design and Oversight
We conducted an international, investigator-ini-
tiated, parallel-group, randomized, open, blinded-
outcome-assessment trial. The details of the trial
design have been published previously.19 The trial
protocol and statistical analysis plan are avail-
able with the full text of this article at NEJM.org. Trial Intervention
The trial was designed and overseen by an execu- Treatment of cardiovascular conditions, anticoagu-
tive committee supported by a steering commit- lation, and rate control were mandated in all pa-
tee and was conducted in accordance with the
Declaration of Helsinki and the International
Council for Harmonisation Good Clinical Prac-
tice Guidelines. An independent data and safety
monitoring board guided the trial. All serious
adverse events were adjudicated by an indepen-
dent end-point review committee, the members
of which were not aware of the treatment-group
assignments. The trial was planned by the Atrial
Fibrillation Network (AFNET) and the European
Heart Rhythm Association. AFNET was respon-
sible for the conduct of the trial. The protocol was
approved by the ethics review boards of all the
institutions involved. Written informed consent
was provided by all patients who participated in
the trial.
AFNET conducted the trial while being ad-
vised by the trial committees and working with
the contract research organization CRI — the
Clinical Research Institute. The contract research
organization and the study sites used the Marvin
electronic case-report form system (XClinical).
The Institute of Medical Biometry and Epidemi-
ology at the University Medical Center Hamburg–
Eppendorf served as the core statistical unit. The
funders of the trial did not influence the trial
design, data collection, analysis, or the decision
to publish. The first author wrote the first draft
of the manuscript. All voting members of the
executive steering committee (see the Supplemen-
tary Appendix, available at NEJM.org) vouch for
the accuracy and completeness of the data and for
the fidelity of the trial to the protocol.
Trial Population
We enrolled adults (≥18 years of age) who had
early atrial fibrillation (defined as atrial fibrilla-
tion diagnosed ≤12 months before enrollment)
and who were older than 75 years of age, had
had a previous transient ischemic attack or stroke,
or met two of the following criteria: age greater
than 65 years, female sex, heart failure, hyperten-
sion, diabetes mellitus, severe coronary artery dis-
ease, chronic kidney disease (Modification of Diet
in Renal Disease stage 3 or 4 [glomerular filtra-
tion rate, 15 to 59 ml per minute per 1.73 m2 of
body-surface area]), and left ventricular hyper-
trophy (diastolic septal wall width, >15 mm).

2 n engl j med  nejm.org

The New England Journal of Medicine

Downloaded from nejm.org by gaspar caponi on September 2, 2020. For personal use only. No other uses without permission.
Copyright © 2020 Massachusetts Medical Society. All rights reserved.
 Early Rhythm-Control Ther apy
tients, in accordance with guideline recommen-
dations.20-22 Patients were randomly assigned in
a 1:1 ratio to receive early rhythm control or usual
care, with randomization stratified according to
site and with variable block lengths used for con-
cealment of assignments.
Early rhythm control required antiarrhythmic
drugs or atrial fibrillation ablation, as well as
cardioversion of persistent atrial fibrillation, to
be initiated early after randomization. Local study
teams chose the type of rhythm-control therapy
independently to deliver this treatment, using
protocol guidance based on current guidelines.20-22
Patients who were randomly assigned to early
rhythm-control therapy were asked to transmit a
patient-operated single-lead electrocardiogram
(ECG) (Vitaphone) twice per week and when symp-
tomatic. All abnormal ECG recordings were
forwarded to the study site. Documentation of re-
current atrial fibrillation triggered an in-person
visit from the site team to escalate rhythm-con-
trol therapy as clinically indicated.
Patients who were randomly assigned to usual
care were initially treated with rate-control ther-
apy without rhythm-control therapy. Rhythm-con-
trol therapy was used only to mitigate uncon-
trolled atrial fibrillation–related symptoms during
adequate rate-control therapy (i.e., therapy that
maintained the heart rate within guideline-rec-
ommended targets).
Outcomes and Adverse Events
The first primary outcome was a composite of
death from cardiovascular causes, stroke (either
ischemic and hemorrhagic), or hospitalization
with worsening of heart failure or acute coronary
syndrome, analyzed in a time-to-event analysis.
The second primary outcome was the number of
nights spent in the hospital per year. Secondary
outcomes reported here include each component
of the first primary outcome (analyzed in a time-
to-event analysis), rhythm, left ventricular func-
tion, quality of life (assessed with the European
Quality of Life–5 Dimensions [EQ-5D] visual
analogue scale and the 12-Item Short-Form Gen-
eral Health Survey [SF-12]), atrial fibrillation–
related symptoms (assessed as the European
Heart Rhythm Association [EHRA] score), and
cognitive function (based on the Montreal Cog-
nitive Assessment [MoCA]) at 2 years. All the
secondary outcomes are listed in Table S5 in the
Supplementary Appendix.
n engl j med  nejm.org
The New England Journal of Medicine
Downloaded from nejm.org by gaspar caponi on September 2, 2020. For personal use only. No other uses without permission.
Copyright © 2020 Massachusetts Medical Society. All rights reserved.
The primary safety outcome was a composite
of death from any cause, stroke, or prespecified
serious adverse events of special interest cap-
turing complications of rhythm-control therapy.
Source data on all potential serious adverse events
and adverse events of special interest were cen-
trally adjudicated by the end-point review com-
mittee.
Follow-up
All patients remained in follow-up from ran-
domization until the end of the trial, death, or
withdrawal from the trial. At baseline, a medical
history; information on clinical characteristics,
therapy, and symptom status (EHRA); responses
on the MoCA, EQ-5D, and SF-12 questionnaires;
and an ECG and echocardiogram were obtained.
A blood specimen was obtained from patients
who consented to participate in a biomarker sub-
study. Every 6 months, trial sites mailed question-
naires to all patients to obtain information on
hospitalizations and cardiovascular events. Ques-
tionnaires were reviewed at the contract research
organization, and source documents for all pos-
sible events were requested from the sites. At
1 and 2 years, an in-person interview, physical
examination, and ECG were performed. The
MoCA, EQ-5D, SF-12, and echocardiography were
repeated at 2 years.
Statistical Analysis
The trial was designed as an event-driven trial.
The first and second primary outcomes were
tested independently for differences between the
treatment groups at an overall two-sided type 1
error rate of 4% for the first primary outcome
and 1% for the second primary outcome to reach
an overall type 1 error rate of 5%. A between-
group difference of 20% in the annual rate of the
first primary outcome was deemed a clinically
relevant difference. We calculated that 685 events
would be needed to show a 20% difference in the
event rate for the first primary outcome with a
power of 80%.
Under the assumption of an event rate of 8%
per year in the control group, a recruitment time
of 48 months, a minimum follow-up time of 24
months, and a loss-to-follow-up of 5% of the ob-
servation time, a sample of 2810 patients was cal-
culated to be needed. After a prespecified blinded
interim analysis of pooled event data that was
performed after 42 months of recruitment, fol-
3
 The n e w e ng l a n d j o u r na l
low-up time was increased to 30 months and the
recruitment period to 65 months, resulting in a
modified sample of 2745 patients without modi-
fying the required number of events. Three un-
blinded interim analyses for early determination
of significance were conducted by the data and
safety monitoring board when 25%, 50%, and
75% of the required events of the first primary
outcome had occurred.
The analyses of the primary outcomes includ-
ed all patients who underwent randomization and
at least one follow-up assessment. The analysis
of the first primary outcome was a comparison
of end-point review committee–adjudicated events
between the treatment groups. The analysis fol-
lowed a group-sequential design with three in-
terim analyses with O’Brien–Fleming stopping
boundaries and two-sided log-rank tests com-
paring early rhythm control with usual care.
Deaths from noncardiovascular causes were treat-
ed as censored. Additional events at the termina-
tion of the trial were included with the use of the
inverse normal method.23 As the primary result of
the trial, the two-sided P value based on Tsiatis,
Rosner, and Mehta stagewise ordering, accom-
panied by the corresponding median unbiased
estimate of the hazard ratio and 96% confidence
interval, is given.24
The second primary outcome was calculated
as the observed sum of nights in the hospital
divided by the individual follow-up time (in days;
in the case of a follow-up time of 0 days, 0.01
days of follow-up was assumed) and reported as
annualized rates. The difference between the
treatment groups was estimated as the arithme-
tic mean and t-based 99% confidence interval.
For the primary analysis of the second primary
outcome, a mixed negative binomial regression
model was used. Explanations of the sensitivity
analyses and analyses of secondary outcomes
and further statistical details are provided in the
Supplementary Appendix.
We used a multiple-imputation procedure with
60 imputations to replace missing values for con-
tinuous outcomes and covariates defined for ad-
justment. With the exception of the primary analy-
sis, estimates are reported with two-sided 95%
confidence intervals throughout (see the statistical
analysis plan in the protocol). These confidence
intervals have not been adjusted for multiplicity
and cannot be used to infer treatment effects. All
analyses were conducted with Stata software, ver-
4 n engl j med  nejm.org
The New England Journal of Medicine
Downloaded from nejm.org by gaspar caponi on September 2, 2020. For personal use only. No other uses without permission.
Copyright © 2020 Massachusetts Medical Society. All rights reserved.
of m e dic i n e
sion 16.1 (StataCorp), and R software, version 3.6.1
(R Project for Statistical Computing).
R e sult s
Trial Participants
A total of 2789 patients underwent randomiza-
tion across 135 sites in 11 European countries
between July 28, 2011, and December 30, 2016.
The primary intention-to-treat population con-
sisted of all 2789 patients — 1395 assigned to
early rhythm control and 1394 assigned to usual
care (Fig. 1). Most patients received guideline-
recommended anticoagulation and therapy for
cardiovascular conditions (Table 1). Patients
were enrolled a median of 36 days (interquartile
range, 6 to 112) after the first diagnosis of atrial
fibrillation. Demographic and clinical character-
istics were generally well balanced between the
groups, although the use of digitalis glycosides
and beta-blockers was slightly more common
(probably because of the group assignment), and
statin use slightly less common, among the pa-
tients assigned to usual care (Table 1, Fig. 1, and
Table S3).
Intervention
Almost all patients (1323 [94.8%]) who were ran-
domly assigned to early rhythm control received
an antiarrhythmic drug or underwent atrial fibril-
lation ablation (Fig. 1), which replicated clinical
practice patterns.25 Among the 1395 patients,
216 (15.4%) had a triggered visit to adapt rhythm-
control therapy. At 2 years, 908 of 1395 patients
(65.1%) were still receiving rhythm-control ther-
apy (Fig. 1).
Usual care consisted of treatment with rate-
control therapy without rhythm-control therapy
throughout follow-up in the majority of patients
assigned to this group. Initially, 1335 (95.8%) of
the 1394 patients in this group had their condi-
tion managed without rhythm-control therapy; at
2 years, 1191 of the 1394 patients (85.4%) were
still not receiving rhythm-control therapy (Fig. 1).
Sinus rhythm was found more often in pa-
tients who had been randomly assigned to receive
early rhythm control (84.9% at 1 year, 82.1% at 2
years) than in patients assigned to receive usual
care (65.5% at 1 year, 60.5% at 2 years) (Table 2;
imputed estimates are provided in Fig. S3). At
2 years, 1020 of 1159 patients (88.0%) assigned
to early rhythm control and 1065 of 1171 patients
 Early Rhythm-Control Ther apy

(90.9%) assigned to usual care were still taking years per patient. A first-primary-outcome event
oral anticoagulants. occurred in 249 patients assigned to receive early
rhythm control (3.9 per 100 person-years) and in
Primary Outcomes 316 patients assigned to receive usual care (5.0 per
The trial was stopped for efficacy at the third 100 person-years) (Table 2). When the results were
interim analysis after a median follow-up of 5.1 adjusted for the group-sequential design of the

2810 Patients were assessed for eligibility

21 Did not meet inclusion

criteria

2789 Underwent randomization

at 135 sites in 11 countries

1395 Were assigned to early

1394 Were assigned to usual care
rhythm control

1395 Were included in primary analysis 1394 Were included in primary analysis

Rhythm Control Chosen by Site

Initial 2 Yr Initial 2 Yr
100 72 (5.2%) 100 100 100
106 (7.6%)
97 (7.0%) 487 (34.9%)
80 80 80 80
None
45
501 (35.9%) (3.2%) Other antiarrhythmic
60 60 53 (3.8%) 60 60
Percent

Percent

1191 (85.4%) drug

1335 (95.8%)
293 (21.0%) Propafenone
40 40 40 40
274 (19.6%) Flecainide
164 (11.8%)
40
83 (5.9%) (2.9%)
Amiodarone
20 233 (16.7%)
20 20 20
39 Dronedarone
270 (19.4%) 28 (2.8%)
112 (8.0%) (2.0%) 97 (7.0%) AF ablation
0 0 0 0

Total follow-up yr lost: 681/7596 (9.0%)
524 (6.9%) Follow-up yr lost because
123 Withdrew
157 (2.1%) Follow-up yr lost because
102 Were lost to follow-up
Total follow-up yr lost: 491/7479 (6.6%)
339 (4.5%) Follow-up yr lost because
83 Withdrew
152 (2.0%) Follow-up yr lost because
106 Were lost to follow-up

Figure 1. Screening, Randomization, Treatment, and Follow-up.

Most of the patients assigned to early rhythm-control therapy were initially treated with antiarrhythmic drugs, often
flecainide. After 2 years of follow-up, 908 of the patients (65.1%) who had been randomly assigned to early rhythm-
control therapy were still receiving active rhythm-control therapy (270 patients treated with atrial fibrillation [AF] ab-
lation and 638 treated with antiarrhythmic drugs), and only 203 patients (14.6%) who had been randomly assigned
to usual care were receiving rhythm-control therapy (97 treated with AF ablation and 106 treated with antiarrhythmic
drugs). All patients who underwent randomization were included in the primary analysis.

n engl j med  nejm.org 5

The New England Journal of Medicine
Downloaded from nejm.org by gaspar caponi on September 2, 2020. For personal use only. No other uses without permission.
Copyright © 2020 Massachusetts Medical Society. All rights reserved.
 The n e w e ng l a n d j o u r na l of m e dic i n e

Table 1. Demographic and Clinical Characteristics of the Patients at Baseline.*

Early Rhythm Control Usual Care

Characteristic (N = 1395) (N = 1394)

Age — yr 70.2±8.4 70.4±8.2

Female sex — no. (%) 645 (46.2) 648 (46.5)
Body-mass index† 29.2±5.4 29.3±5.4
Type of atrial fibrillation — no./total no. (%)
First episode 528/1391 (38.0) 520/1394 (37.3)
Paroxysmal 501/1391 (36.0) 493/1394 (35.4)
Persistent 362/1391 (26.0) 381/1394 (27.3)
Sinus rhythm at baseline — no./total no. (%) 762/1389 (54.9) 743/1393 (53.3)
Median days since atrial fibrillation diagnosis (IQR)‡ 36.0 (6.0–114.0) 36.0 (6.0–112.0)
Absence of atrial fibrillation symptoms — no./total no. (%)§ 395/1305 (30.3) 406/1328 (30.6)
Previous cardioversion — no./total no. (%) 546/1364 (40.0) 543/1389 (39.1)
Concomitant cardiovascular conditions
Previous stroke or transient ischemic attack — no. (%) 175 (12.5) 153 (11.0)
At least mild cognitive impairment — no./total no. (%)¶ 582/1326 (43.9) 584/1341 (43.5)
Arterial hypertension — no. (%) 1230 (88.2) 1220 (87.5)
Blood pressure — mm Hg‖
Systolic 136.5±19.4 137.5±19.3
Diastolic 80.9±12.1 81.3±12.0
Stable heart failure — no. (%)** 396 (28.4) 402 (28.8)
CHA2DS2-VASc score†† 3.4±1.3 3.3±1.3
Valvular heart disease — no./total no. (%) 609/1389 (43.8) 642/1391 (46.2)
Chronic kidney disease of MDRD stage 3 or 4 — no. (%)‡‡ 172 (12.3) 179 (12.8)
Medication at discharge — no./total no. (%)§§
Oral anticoagulation with NOAC or VKA 1267/1389 (91.2) 1250/1393 (89.7)
Digoxin or digitoxin 46/1389 (3.3) 85/1393 (6.1)
Beta-blocker 1058/1389 (76.2) 1191/1393 (85.5)
ACE inhibitors or angiotensin II receptor blocker 953/1389 (68.6) 979/1393 (70.3)
Mineralocorticoid-receptor antagonist 90/1389 (6.5) 92/1393 (6.6)
Diuretic 559/1389 (40.2) 561/1393 (40.3)
Statin 628/1389 (45.2) 568/1393 (40.8)
Platelet inhibitor 229/1389 (16.5) 226/1393 (16.2)

* Plus–minus values are means ±SD. Definitions of clinical measures are provided in Table S1. ACE denotes angiotensin-converting en-
zyme, IQR interquartile range, NOAC non–vitamin K antagonist oral anticoagulant, and VKA vitamin K antagonist.
† The body-mass index is the weight in kilograms divided by the square of the height in meters. Data were missing for 7 patients assigned
to early rhythm control and for 6 patients assigned to usual care.
‡ Data on median days since atrial fibrillation diagnosis were missing for 2 patients assigned to early rhythm control and for 1 patient as-
signed to usual care.
§ The absence of symptoms was defined as a European Heart Rhythm Association (EHRA) score of I. The EHRA score categorizes symp-
toms related to atrial fibrillation into four classes from I (asymptomatic) to IV (severe symptoms at rest).
¶ At least mild cognitive impairment was defined as a Montreal Cognitive Assessment (MoCA) score of less than 26. The MoCA score pro-
vides an overall assessment of cognitive function. Scores range from 0 to 30, with lower scores indicating worse cognitive function.
‖ Data on blood pressure were missing for 9 patients assigned to early rhythm control and 4 patients assigned to usual care.
** Stable heart failure was defined as New York Heart Association stage II or a left ventricular ejection fraction of less than 50%.
†† CHA2DS2-VASc scores (an assessment of the risk of stroke among patients with atrial fibrillation) range from 0 to 9, with higher scores
indicating a higher risk of stroke.
‡‡ A Modification of Diet in Renal Disease (MDRD) stage of 3 or 4 indicates a glomerular filtration rate of 15 to 59 ml per minute per 1.73 m2 of
body-surface area.
§§ Because of the high proportion of patients with atrial fibrillation that was first diagnosed at enrollment, important therapies were initiated
between enrollment and discharge from the baseline visit. Therefore, medication at discharge from the baseline visit is shown.

6 n engl j med  nejm.org

The New England Journal of Medicine

Downloaded from nejm.org by gaspar caponi on September 2, 2020. For personal use only. No other uses without permission.
Copyright © 2020 Massachusetts Medical Society. All rights reserved.
 Early Rhythm-Control Ther apy

Table 2. Efficacy Outcomes.*

Outcome Early Rhythm Control Usual Care Treatment Effect

First primary outcome — events/person-yr (incidence/ 249/6399 (3.9) 316/6332 (5.0) 0.79 (0.66 to 0.94)†
100 person-yr)
Components of first primary outcome — events/person-yr
(incidence/100 person-yr)
Death from cardiovascular causes 67/6915 (1.0) 94/6988 (1.3) 0.72 (0.52 to 0.98)‡
Stroke 40/6813 (0.6) 62/6856 (0.9) 0.65 (0.44 to 0.97)‡
Hospitalization with worsening of heart failure 139/6620 (2.1) 169/6558 (2.6) 0.81 (0.65 to 1.02)‡
Hospitalization with acute coronary syndrome 53/6762 (0.8) 65/6816 (1.0) 0.83 (0.58 to 1.19)‡
Second primary outcome — nights spent in hospital/yr 5.8±21.9 5.1±15.5 1.08 (0.92 to 1.28)§
Key secondary outcomes at 2 yr
Change in left ventricular ejection fraction — % 1.5±9.8 0.8±9.8 0.23 (−0.46 to −0.91)¶
Change in EQ-5D score‖ −1.0±21.4 −2.7±22.3 1.07 (−0.68 to 2.82)¶
Change in SF-12 Mental Score** 0.7±10.6 1.6±10.1 −1.20 (−2.04 to −0.37)¶
Change in SF-12 Physical Score** 0.3±8.5 0.1±8.2 0.33 (−0.39 to 1.06)¶
Change in MoCA score 0.1±3.3 0.1±3.2 −0.14 (−0.39 to 0.12)¶
Sinus rhythm — no. of patients with feature/total no. (%) 921/1122 (82.1) 687/1135 (60.5) 3.13 (2.55 to 3.84)††
Asymptomatic — no. of patients with feature/total no. (%)‡‡ 861/1159 (74.3) 850/1171 (72.6) 1.14 (0.93 to 1.40)††

* Plus–minus values are means ±SD. Data in columns 2 and 3 are observed data, and data in column 4 are model-based effect estimates.
There were no significant differences in the key secondary outcomes between the treatment groups, with two exceptions: more patients
assigned to early rhythm control were in sinus rhythm at 2 years, and a slightly greater improvement in the 12-Item Short-Form Health
Survey (SF-12) mental score at 2 years was found in the group assigned to usual care. All 95% confidence intervals for secondary end
points were not adjusted for multiplicity and should not be used to infer definitive treatment effects. The results for additional secondary
outcomes are provided in the Supplementary Appendix.
† The treatment effect is expressed as the median unbiased estimate of the hazard ratio and 96% confidence interval, which were calculated
on the basis of Tsiatis, Rosner, and Mehta stagewise ordering that adjusts for the group-sequential design.24 P = 0.005 for the between-
group comparison.
‡ The treatment effect is expressed as the hazard ratio and 95% confidence interval, which were calculated with a Cox regression with treat-
ment group as the fixed factor and site as the shared frailty term.
§ The treatment effect is expressed as the incidence rate ratio and 99% confidence interval, which were calculated with a mixed negative bi-
nomial model with treatment group as the fixed factor, the log of follow-up time as the offset, and site as a random effect. P = 0.23 for the
between-group comparison.
¶ The treatment effect is expressed as the adjusted mean difference and 95% confidence interval, which were calculated with a mixed linear
model with the corresponding baseline measurement and treatment group as the fixed effects and site as a random effect, analyzed after
multiple imputation of missing values in survivors.
‖ The European Quality of Life–5 Dimensions (EQ-5D) assesses state of health on visual analogue scale from 0 (very bad health) to 100
(perfect health); values were defined as 0 for nonsurvivors.
** Scores on the SF-12 range from 0 to 100, with lower scores indicating worse functioning.
†† The treatment effect is expressed as the odds ratio and 95% confidence interval, which were calculated with a mixed logistic regression
including treatment group and the corresponding baseline assessment as fixed factors and site as a random effect, analyzed after multiple
imputation of missing values in survivors.
‡‡ The absence of symptoms was defined as an EHRA score of I.

trial, a first-primary-outcome event was found
to have occurred less often in patients assigned
to early rhythm control than in patients as-
signed to usual care (hazard ratio, 0.79; 96%
confidence interval [CI], 0.66 to 0.94; P = 0.005)
(Fig. 2). The effects of early rhythm control on
individual components of the first primary out-
come were consistent with the overall result
(Table 2 and Fig. S4). The effect of early rhythm
control on the first primary outcome remained
stable after adjustment for relevant covariates
(hazard ratio, 0.78; 95% CI, 0.66 to 0.92; P = 0.004)
(Fig. S1), and the effect was consistent across
subgroups (Fig. S5). There was no significant
difference in the mean (±SD) number of nights
spent in the hospital between the treatment
groups (early rhythm control, 5.8±21.9 days per
year; usual care, 5.1±15.5 days per year; P = 0.23)
(Table 2).
The numbers of patients with a primary-safety-

n engl j med  nejm.org 7

The New England Journal of Medicine
Downloaded from nejm.org by gaspar caponi on September 2, 2020. For personal use only. No other uses without permission.
Copyright © 2020 Massachusetts Medical Society. All rights reserved.
 The n e w e ng l a n d j o u r na l of m e dic i n e

Discussion
100

90 In this multicenter randomized trial, a strategy of

80 initiating rhythm-control therapy in all patients
with early atrial fibrillation and concomitant
Cumulative Incidence (%)

70
60
50
40
30 Usual care
20
Early rhythm control
10
0 of an appreciable difference in hospital nights is
0 2 4 6
Years since Randomization
No. at Risk
Usual care 1394 1169 888 405
Early rhythm control 1395 1193 913 404
Figure 2. Aalen–Johansen Cumulative-Incidence Curves for the First
Primary Outcome.
The first primary outcome was a composite of death from cardiovascular
causes, stroke, or hospitalization with worsening of heart failure or acute
coronary syndrome.
outcome event did not differ significantly be-
tween the treatment groups (early rhythm con-
trol, 231 patients; usual care, 223 patients) (Table 3
and Table S4). Mortality was similar in the two
treatment groups, and stroke occurred less fre-
quently among patients assigned to early rhythm
control than among those assigned to usual care.
Serious adverse events related to rhythm-control
therapy were more common in the group assigned
to early rhythm control but were infrequent; dur-
ing the 5-year follow-up period, such events oc-
curred in 68 patients (4.9%) assigned to early
rhythm control and 19 patients (1.4%) assigned
to usual care (Table 3 and Table S4).
Secondary Outcomes
Left ventricular function and cognitive function
were stable at 2 years, with no evidence of sig-
nificant differences between the treatment
groups (Table 2). Most patients in both groups
were free from atrial fibrillation–related symp-
toms at 2 years, and the change from baseline in
atrial fibrillation–related symptoms (EHRA score)
and quality of life (EQ-5D score) did not differ
significantly between the groups (Table 2).
cardiovascular conditions was associated with a
lower risk of death from cardiovascular causes,
stroke, or hospitalization for heart failure or acute
coronary syndrome than usual care over a follow-
up time of more than 5 years (absolute difference
in risk, 1.1 events per 100 person-years).
Early rhythm control did not affect the num-
ber of nights spent in the hospital. The absence
8 reassuring in view of the excess hospitalizations
associated with rhythm-control therapy reported
in two previous large trials.8,13
34
26
Most patients (>70%) were asymptomatic at
1 and 2 years in both treatment groups, and the
magnitude of change in left ventricular function
did not differ between the groups at 2 years,
which indicates that both rate control and rhythm
control can control symptoms and maintain car-
diac function in patients with early atrial fibril-
lation. The effects of an early rhythm-control
strategy on the primary outcome appeared to be
generally consistent across predefined subgroups,
including asymptomatic patients, patients with
obesity, and patients with or without heart failure.
Previous studies comparing rate-control and
rhythm-control strategies did not show better
outcomes with rhythm control than with rate
control.7,8,12,13 In contrast to those trials, our trial
included atrial fibrillation ablation, a powerful
rhythm-control therapy5,26 that works synergisti-
cally with antiarrhythmic drugs.27,28 It is conceiv-
able that atrial fibrillation ablation contributed
to the superiority of early rhythm control in our
trial. Also, unlike patients in previous tri-
als,7,8,12,13 most patients in both treatment groups
in our trial continued to receive anticoagulation,
rate control, and treatment of concomitant cardio-
vascular conditions, maintaining their protective
effects.
Whereas previous trials have evaluated rhythm
control in patients with established, long-stand-
ing atrial fibrillation,7,8,12,13 we enrolled patients
with early atrial fibrillation and initiated rhythm-
control therapy shortly after the diagnosis of
atrial fibrillation. Furthermore, 54% of the pa-
tients were in sinus rhythm at enrollment. In one
large previous trial, rhythm-control therapy with

8 n engl j med  nejm.org

The New England Journal of Medicine

Downloaded from nejm.org by gaspar caponi on September 2, 2020. For personal use only. No other uses without permission.
Copyright © 2020 Massachusetts Medical Society. All rights reserved.
 Early Rhythm-Control Ther apy

Table 3. Safety Outcomes.*

Early Rhythm Control Usual Care

Outcome (N = 1395) (N = 1394)

number (percent)
Primary composite safety outcome 231 (16.6) 223 (16.0)
Stroke 40 (2.9) 62 (4.4)
Death 138 (9.9) 164 (11.8)
Serious adverse event of special interest related to rhythm-control therapy 68 (4.9) 19 (1.4)
Serious adverse event related to antiarrhythmic drug therapy
Nonfatal cardiac arrest 1 (0.1) 1 (0.1)
Toxic effects of atrial fibrillation–related drug therapy 10 (0.7) 3 (0.2)
Drug-induced bradycardia 14 (1.0) 5 (0.4)
Atrioventricular block 2 (0.1) 0
Torsades de pointes tachycardia 1 (0.1) 0
Serious adverse event related to atrial fibrillation ablation
Pericardial tamponade 3 (0.2) 0
Major bleeding related to atrial fibrillation ablation 6 (0.4) 0
Nonmajor bleeding related to atrial fibrillation ablation 1 (0.1) 2 (0.1)
Other serious adverse event of special interest related to rhythm-control therapy
Blood pressure–related event† 1 (0.1) 0
Hospitalization for atrial fibrillation 11 (0.8) 3 (0.2)
Other cardiovascular event 5 (0.4) 1 (0.1)
Other event 1 (0.1) 3 (0.2)
Syncope 4 (0.3) 1 (0.1)
Hospitalization for worsening of heart failure with decompensated heart failure 3 (0.2) 0
Implantation of a pacemaker, defibrillator, cardiac resynchronization device, or any 8 (0.6) 4 (0.3)
other cardiac device

* Patients could have had more than one event, and therefore the total sum of events is higher than the number of patients with events. For
dichotomous outcomes, mixed logistic-regression models with a random effect for site were used for comparison of random groups. Stroke
was significantly less frequent (P = 0.03) and serious adverse events of special interest significantly more frequent (P<0.001) in the group as-
signed to early rhythm control; the other safety outcomes did not differ significantly between the groups.
† Blood pressure–related events included hypotension and hypertension (excluding syncope).

the antiarrhythmic drug dronedarone was found
to reduce the risk of death or hospitalization for
cardiovascular causes6 and, in a post hoc analy-
sis, was found to reduce the risk of stroke.29 The
majority of patients in that trial had had atrial
fibrillation for less than 1 year (68% of the 2859
patients in whom the duration of atrial fibrilla-
tion was known), and 75% of the patients were
in sinus rhythm at enrollment. Almost no pa-
tients were in sinus rhythm at the time of enroll-
ment in another trial, in which harm was shown
when dronedarone was tested in patients with
chronic atrial fibrillation (most of whom had
had atrial fibrillation for >2 years at enrollment).30
Our results, together with other published evi-
dence, suggest that the early initiation of rhythm-
control therapy probably contributed to the clini-
cal superiority of this strategy.
Early rhythm-control therapy used in the
present trial included all major antiarrhythmic
drugs and atrial fibrillation ablation, and there
were no significant differences between the treat-
ment groups with respect to the primary safety
outcome. Early rhythm control was associated
with more adverse events related to rhythm-con-
trol therapy than was usual care, but such events
were uncommon, similar to the results of other
recent trials comparing rhythm-control therapies
in patients with atrial fibrillation.5,26 Early initia-
tion of rhythm-control therapy, guidance on the

n engl j med  nejm.org 9

The New England Journal of Medicine
Downloaded from nejm.org by gaspar caponi on September 2, 2020. For personal use only. No other uses without permission.
Copyright © 2020 Massachusetts Medical Society. All rights reserved.
 The n e w e ng l a n d j o u r na l of m e dic i n e

safe use of antiarrhythmic drugs,20-22 and the
availability of atrial fibrillation ablation may have
contributed to the low incidence of adverse events
associated with rhythm-control therapy, as com-
pared with previous trials.7,8,12,13
Some limitations of our trial should be noted.
We compared two treatment strategies that ne-
cessitated an open trial design. Blinded, central
assessment of primary outcomes was used to
minimize bias. The trial was not primarily de-
signed to assess the safety and effectiveness of
specific components of early rhythm control. We
enrolled only patients with early atrial fibrilla-
tion, and thus the results may not be generaliz-
able to patients in whom rhythm-control therapy
that includes atrial fibrillation ablation is initi-
ated later. Further analysis is needed of the costs
of early rhythm control. All enrolled patients were
deemed eligible for either rate-control or rhythm-
control therapy, which probably excluded the most
symptomatic patients. We did not collect detailed
information on recurrent atrial fibrillation in
both groups, and therefore our data on percent-
ages of patients with sinus rhythm are not com-
parable to data on recurrent atrial fibrillation
from other rhythm-control trials.5,11,26
Early initiation of rhythm-control therapy was
associated with less frequent cardiovascular events
than usual care in patients with early atrial fi-
brillation and cardiovascular conditions without
affecting the number of nights spent in the hos-
pital. As expected, the early rhythm-control strat-
egy was associated with more adverse events re-
lated to rhythm-control therapy, but the incidence
of the overall safety outcome events was similar
in the two groups. These results are relevant to
decisions regarding rhythm-control therapy in pa-
tients with early atrial fibrillation.
Supported by a grant from the German Ministry of Educa-
tion and Research (01 GI 0204), the German Center for Car-
diovascular Research (DZHK), the Atrial Fibrillation Network
(AFNET), the European Heart Rhythm Association, St. Jude
Medical–Abbott, Sanofi, the German Heart Foundation, the
European Union (grant agreement 633196 [CATCH ME], to
Dr. Kirchhof and AFNET and grant agreement EU IMI 116074
[BigData@Heart], to Dr. Kirchhof), the British Heart Foun-
dation (FS/13/43/30324, PG/17/30/32961, PG/20/22/35093, and
AA/18/2/34218, to Dr. Kirchhof), and the Leducq Foundation
(to Dr. Kirchhof).
Disclosure forms provided by the authors are available with
the full text of this article at NEJM.org.
A data sharing statement provided by the authors is available
with the full text of this article at NEJM.org.
We thank all the patients who agreed to participate in the
trial, and especially those who stayed in the trial for many years;
all local study teams; the dedicated staff at AFNET and CRI —
the Clinical Research Institute; and all committee members (see
the Supplementary Appendix).

Appendix
The authors’ full names and academic degrees are as follows: Paulus Kirchhof, M.D., A. John Camm, M.D., Andreas Goette, M.D., Axel
Brandes, M.D., Lars Eckardt, M.D., Arif Elvan, M.D., Thomas Fetsch, M.D., Isabelle C. van Gelder, M.D., Doreen Haase, Ph.D., Lau-
rent M. Haegeli, M.D., Frank Hamann, M.D., Hein Heidbüchel, M.D., Ph.D., Gerhard Hindricks, M.D., Josef Kautzner, M.D., Karl‑Heinz
Kuck, M.D., Lluis Mont, M.D., G. Andre Ng, M.B., Ch.B., Ph.D., Jerzy Rekosz, M.D., Norbert Schoen, M.D., Ulrich Schotten, M.D.,
Ph.D., Anna Suling, Ph.D., Jens Taggeselle, M.D., Sakis Themistoclakis, M.D., Eik Vettorazzi, M.Sc., Panos Vardas, M.D., Ph.D., Karl
Wegscheider, Ph.D., Stephan Willems, M.D., Harry J.G.M. Crijns, M.D., Ph.D., and Günter Breithardt, M.D.
The authors’ affiliations are as follows: the Department of Cardiology, University Heart and Vascular Center (P.K.), and Institute of
Medical Biometry and Epidemiology (A.S., E.V., K.W.), University Medical Center Hamburg–Eppendorf, LANS Cardio (K.-H.K.), and
the Department of Cardiology, Asklepios Klinik St. Georg (S.W.), Hamburg, Atrial Fibrillation Network (AFNET) (P.K., A.G., L.E., T.F.,
D.H., K.-H.K., N.S., U.S., J.T., K.W., S.W., G.B.) and the Department of Cardiology II (Electrophysiology), University Hospital Münster
(L.E., G.B.), Münster, the German Center of Cardiovascular Research, Partner Site Hamburg/Lübeck/Kiel (P.K., K.W., S.W.), St. Vincenz
Hospital, Paderborn (A.G.), the Working Group of Molecular Electrophysiology, University Hospital Magdeburg, Magdeburg (A.G.), the
Clinical Research Institute, Munich (T.F.), Hospital Konstanz, Konstanz (F.H.), the Department of Cardiology and Electrophysiology,
University Heart Center–Helios, and Leipzig Heart Institute, Leipzig (G.H.), University Heart Center Schleswig–Holstein, Campus Lü-
beck, Lübeck (K.-H.K.), Cardiology Practice Schön, Mühldorf (N.S.), and Cardiology Practice Taggeselle, Markkleeberg (J.T.) — all in
Germany; the Institute of Cardiovascular Sciences, University of Birmingham, Birmingham (P.K.), the Cardiology Clinical Academic
Group, Molecular and Clinical Sciences Research Institute, St. George’s University of London, London (A.J.C.), and the Department of
Cardiovascular Sciences, University of Leicester, National Institute for Health Research Leicester Biomedical Research Centre, Glenfield
Hospital, Leicester (G.A.N.) — all in the United Kingdom; the Department of Cardiology, Odense University Hospital, and Department
of Clinical Research, University of Southern Denmark, Odense (A.B.); Isala Hospital and Diagram B.V., Zwolle (A.E.), the University of
Groningen, University Medical Center Groningen, Groningen (I.C.G.), and the Department of Physiology, Cardiovascular Research
Institute Maastricht (U.S.), and the Department of Cardiology, Maastricht University Medical Center and Cardiovascular Research Insti-
tute Maastricht (H.J.G.M.C.), Maastricht — all in the Netherlands; University Hospital Zurich, Zurich (L.M.H.), and the Division of
Cardiology, Medical University Department, Kantonsspital Aarau, Aarau (L.M.H.) — both in Switzerland; University Hospital Antwerp
and Antwerp University, Antwerp, Belgium (H.H.); the Institute for Clinical and Experimental Medicine, Prague, Czech Republic (J.K.);
the Hospital Clinic, University of Barcelona and Institut de Recerca Biomèdica, August Pi-Sunyer, Barcelona (L.M.), and Centro Inves-
tigación Biomedica en Red Cardiovascular, Madrid (L.M.); Department of Cardiology, Hospital Wojewódzka Stacja Pogotowia Ra-
tunkowego i Transportu Sanitarnego (WSRiTS) Meditrans, Warsaw, Poland (J.R.); the Department of Cardiology, Ospedale dell’Angelo,
Venice, Italy (S.T.); and Heart Sector, Hygeia Hospitals Group, Athens (P.V.).

10 n engl j med  nejm.org

The New England Journal of Medicine

Downloaded from nejm.org by gaspar caponi on September 2, 2020. For personal use only. No other uses without permission.
Copyright © 2020 Massachusetts Medical Society. All rights reserved.
 Early Rhythm-Control Ther apy
References
1. Marijon E, Le Heuzey J-Y, Connolly S,
et al. Causes of death and influencing fac-
tors in patients with atrial fibrillation: a
competing-risk analysis from the Ran-
domized Evaluation of Long-Term Antico-
agulant Therapy Study. Circulation 2013;​
128:​2192-201.
2. Willems S, Meyer C, de Bono J, et al.
Cabins, castles, and constant hearts:
rhythm control therapy in patients with
atrial fibrillation. Eur Heart J 2019;​ 40:​
3793-3799c.
3. Kirchhof P, Radaideh G, Kim YH, et al.
Global prospective safety analysis of riv­a­
roxaban. J Am Coll Cardiol 2018;​72:​141-53.
4. Ruff CT, Giugliano RP, Braunwald E,
et al. Comparison of the efficacy and
safety of new oral anticoagulants with
warfarin in patients with atrial fibrilla-
tion: a meta-analysis of randomised tri-
als. Lancet 2014;​383:​955-62.
5. Packer DL, Mark DB, Robb RA, et al.
Effect of catheter ablation vs antiarrhyth-
mic drug therapy on mortality, stroke,
bleeding, and cardiac arrest among pa-
tients with atrial fibrillation: the CABANA
randomized clinical trial. JAMA 2019;​321:​
1261-74.
6. Hohnloser SH, Crijns HJGM, van
Eickels M, et al. Effect of dronedarone on
cardiovascular events in atrial fibrilla-
tion. N Engl J Med 2009;​360:​668-78.
7. Van Gelder IC, Hagens VE, Bosker
HA, et al. A comparison of rate control and
rhythm control in patients with recurrent
persistent atrial fibrillation. N Engl J Med
2002;​347:​1834-40.
8. Wyse DG, Waldo AL, DiMarco JP, et al.
A comparison of rate control and rhythm
control in patients with atrial fibrillation.
N Engl J Med 2002;​347:​1825-33.
9. Benjamin EJ, Wolf PA, D’Agostino RB,
Silbershatz H, Kannel WB, Levy D. Impact
of atrial fibrillation on the risk of death:
the Framingham Heart Study. Circulation
1998;​98:​946-52.
10. Nattel S, Guasch E, Savelieva I, et al.
Early management of atrial fibrillation to
prevent cardiovascular complications. Eur
Heart J 2014;​35:​1448-56.
11. Kirchhof P, Bax J, Blomstrom-Lund­
quist C, et al. Early and comprehensive
management of atrial fibrillation: execu-
tive summary of the proceedings from the
2nd AFNET-EHRA consensus conference
‘research perspectives in AF.’ Eur Heart J
2009;​30:​2969-77c.
The New England Journal of Medicine
Downloaded from nejm.org by gaspar caponi on September 2, 2020. For personal use only. No other uses without permission.
Copyright © 2020 Massachusetts Medical Society. All rights reserved.
12. Carlsson J, Miketic S, Windeler J, et al.
Randomized trial of rate-control versus
rhythm-control in persistent atrial fibril-
lation: the Strategies of Treatment of
Atrial Fibrillation (STAF) study. J Am Coll
Cardiol 2003;​41:​1690-6.
13. Roy D, Talajic M, Nattel S, et al.
Rhythm control versus rate control for
atrial fibrillation and heart failure. N Engl
J Med 2008;​358:​2667-77.
14. Marrouche NF, Brachmann J, Andre-
sen D, et al. Catheter ablation for atrial
fibrillation with heart failure. N Engl J
Med 2018;​378:​417-27.
15. Tsadok MA, Jackevicius CA, Essebag
V, et al. Rhythm versus rate control thera-
py and subsequent stroke or transient
ischemic attack in patients with atrial fi-
brillation. Circulation 2012;​126:​2680-7.
16. Themistoclakis S, Corrado A, March-
linski FE, et al. The risk of thromboembo-
lism and need for oral anticoagulation
after successful atrial fibrillation abla-
tion. J Am Coll Cardiol 2010;​55:​735-43.
17. Bunch TJ, Crandall BG, Weiss JP, et al.
Patients treated with catheter ablation for
atrial fibrillation have long-term rates of
death, stroke, and dementia similar to pa-
tients without atrial fibrillation. J Cardio-
vasc Electrophysiol 2011;​22:​839-45.
18. Noseworthy PA, Gersh BJ, Kent DM, et
al. Atrial fibrillation ablation in practice:
assessing CABANA generalizability. Eur
Heart J 2019;​40:​1257-64.
19. Kirchhof P, Breithardt G, Camm AJ, et
al. Improving outcomes in patients with
atrial fibrillation: rationale and design of
the Early treatment of Atrial fibrillation
for Stroke prevention Trial. Am Heart J
2013;​166:​442-8.
20. Camm AJ, Kirchhof P, Lip GYH, et al.
Guidelines for the management of atrial
fibrillation: the Task Force for the Man-
agement of Atrial Fibrillation of the Euro-
pean Society of Cardiology (ESC). Euro-
pace 2010;​12:​1360-420.
21. January CT, Wann LS, Alpert JS, et al.
2014 AHA/ACC/HRS guideline for the
management of patients with atrial fibril-
lation: a report of the American College of
Cardiology/American Heart Association
Task Force on Practice Guidelines and the
Heart Rhythm Society. J Am Coll Cardiol
2014;​64(21):​e1-e76.
22. Kirchhof P, Benussi S, Kotecha D, et
al. 2016 ESC guidelines for the manage-
ment of atrial fibrillation developed in
n engl j med  nejm.org
collaboration with EACTS. Eur Heart J
2016;​37:​2893-962.
23. Committee for Medicinal Products for
Human Use (CHMP). Reflection paper on
methodological issues in confirmatory clini-
cal trials planned with an adaptive design.
London:​European Medicines Agency, Octo-
ber 18, 2007 (https://www​.­ema​.­europa​.­eu/​
­en/​­documents/​­scientific​-­g uideline/​
­reflection​-­paper​-­methodological​-­issues​
-­confirmatory​-­clinical​-­t rials​-­planned​
-­adaptive​-­design_en​.­pdf).
24. Tsiatis AA, Rosner GL, Mehta CR. Ex-
act confidence intervals following a group
sequential test. Biometrics 1984;​ 40:​797-
803.
25. Glorioso TJ, Grunwald GK, Ho PM,
Maddox TM. Reference effect measures
for quantifying, comparing and visualiz-
ing variation from random and fixed ef-
fects in non-normal multilevel models,
with applications to site variation in med-
ical procedure use and outcomes. BMC
Med Res Methodol 2018;​18:​74.
26. Cosedis Nielsen J, Johannessen A,
Raatikainen P, et al. Radiofrequency abla-
tion as initial therapy in paroxysmal atrial
fibrillation. N Engl J Med 2012;​367:​1587-
95.
27. Duytschaever M, Demolder A, Phlips
T, et al. PulmOnary vein isolation With vs.
without continued antiarrhythmic Drug
trEatment in subjects with Recurrent Atri-
al Fibrillation (POWDER AF): results from
a multicentre randomized trial. Eur Heart
J 2018;​39:​1429-37.
28. Darkner S, Chen X, Hansen J, et al.
Recurrence of arrhythmia following short-
term oral AMIOdarone after CATheter ab-
lation for atrial fibrillation: a double-blind,
randomized, placebo-controlled study
(AMIO-CAT trial). Eur Heart J 2014;​ 35:​
3356-64.
29. Connolly SJ, Crijns HJGM, Torp-Ped-
ersen C, et al. Analysis of stroke in
ATHENA: a placebo-controlled, double-
blind, parallel-arm trial to assess the ef-
ficacy of dronedarone 400 mg BID for the
prevention of cardiovascular hospitaliza-
tion or death from any cause in patients
with atrial fibrillation/atrial flutter. Cir-
culation 2009;​120:​1174-80.
30. Connolly SJ, Camm AJ, Halperin JL, et
al. Dronedarone in high-risk permanent
atrial fibrillation. N Engl J Med 2011;​365:​
2268-76.
Copyright © 2020 Massachusetts Medical Society.
11